sastpg/CoVo_Dataset

Consistent Paths Lead to Truth: Self-Rewarding Reinforcement Learning for LLM Reasoning The file test.jsonl is used to monitor the model performance as training proceeds. Level Source 1 GSM8K 2 MATH-500 3 AMC-23 4 AIME2024 5 AIME2025